Family Support in Hard Times: Dynamics of IntergenerationalExchange after Adverse Events
Abstract: We use an event-study approach to examine changes in intergenerational financial transfers andinformal care within families following wealth loss, job exit, widowhood, and health shocks. We find sharpreductions in parental giving to adult children following negative shocks to parents’ wealth and earnedincome, particularly in low-wealth households. Parental giving also decreases with some health shocksand increases following spousal death. Meanwhile, children of low-wealth households increase financialtransfers to their parents following adverse shocks and children in both high- and low-wealth householdsincrease their provision of informal care to parents following a wide range of adverse shocks.
Jessamyn Schaller1
Claremont McKenna College and NBER
Robert Day School of Economics and Finance
500 East Ninth Street
Claremont, CA 91711
Email: [email protected]
Chase Eck1
University of Arizona
Department of Economics
1130 E Helen Street Suite 401
Tucson, AZ 85721-0108
Email: [email protected]
1Funding for this project was provided by the W.E. Upjohn Institute for Employment Research and by the Eller College ofManagement at the University of Arizona. An earlier version of the paper is posted as Upjohn Institute working paper 19-313,with the title “Adverse Life Events and Intergenerational Transfers.” We are grateful to Gary Solon, Steven Haider, Hilary Hoynes,Heather Antecol, Itzik Fadlon, Mariana Zerpa, and participants of UC Santa Barbara’s 2019 conference on “Health and LaborMarket Effects of Public Policy” for helpful comments.
1. IntroductionThe consequences of adverse life events in households of retirement age are well documented in a lit-
erature spanning many decades and several disciplines. Researchers have studied the effects of wealth loss,
employment transitions, health shocks, and household structure changes on finances, labor supply, consump-
tion, and health trajectories in older households.1 Because of their prevalence, health shocks in particular
have been the subject of a substantial body of research (see Prinz et al. (2018) for a recent summary). There
has also been growing interest in understanding the effects of wealth loss and job displacement among older
Americans, as research has shown that the wealth and employment trajectories of older households were
substantially and permanently impacted by the severe downturns in the housing, financial, and labor mar-
kets that occurred during the Great Recession of 2007–2009 (Munnell and Rutledge, 2013).2 This area of
research will likely continue to grow in the coming years given the implications of the COVID-19 pandemic
and the associated economic downturn for health and financial stability in older households.
While there is broad interest in the direct effects of adverse life events on older households that experi-
ence them, little attention has been paid to the intergenerational transmission of those effects—how negative
shocks in parents’ households affect the outcomes of adult children. Among other potential family effects
(see, for example, spillovers in health behaviors in Fadlon and Nielsen 2019), adverse events may disrupt
wealth transfers from parents to their children. Disruptions in the flow of inter vivos transfers from parents
to children could have large welfare effects, as transfer amounts can be economically significant3 and are
typically given at times when children’s marginal utility of income is high (McGarry, 2016). The intergener-
ational transmission of adverse shocks may exacerbate the effects of aggregate downturns for young adults,
who are already vulnerable to business cycle fluctuations (Hoynes, Miller and Schaller, 2012).
It is also possible that adult children could play an important role in helping their parents to recover from
adverse events. Research has examined parents’ role in helping children after negative shocks (Edwards,
2019; Kaplan, 2012; McGarry, 2016), but less attention has been paid to children’s role in alleviating the
impacts of negative shocks for their parents. Though upstream financial transfers from children to parents are
generally smaller and less frequent than downstream transfers, they may also have significant welfare effects
if they are given in response to declining income, wealth, or health in parental households. Adult children,
particularly female children and children in minority households, also provide care for aging parents, and
thus may also help to smooth their parents’ consumption by increasing their provision of in-kind assistance,
including helping with everyday activities and health care. Any increases in caregiving following adverse
shocks could also have important implications for the younger generation, as caregiving has been found to
be associated with reduced work productivity and increases in emotional distress (Wolff et al., 2016).
In this paper, we use panel data spanning more than 20 years from the Health and Retirement Study
1See Schwandt (2018) and Pool et al. (2018) on the effects of wealth shocks, Chan and Stevens (2001) and Salm (2009) on theeffects of job loss, Dobkin et al. (2018) and Wu (2003) on the effects of health shocks, and Sharma (2015) and Goda, Shoven andSlavov (2013) on the effects of divorce and widowhood.
2For example, see Farber (2017); Christelis, Georgarakos and Jappelli (2015).3The average transfer recorded in our HRS sample is $13,601.
1
to examine the dynamics of intergenerational transfers between parents and their adult children following
adverse events occurring in parent households. Using an event-study approach, we estimate within-family
changes in financial transfers and informal care following several different life events impacting the older
generation: sudden wealth loss, exit from employment, fatal health shocks (widowhood), and three different
morbidity shocks: hospitalization, disability, and poor health.
Though the events that we consider are not necessarily exogenous, conditional on experiencing a par-
ticular event, event timing is arguably quasi-random and, in many cases, unanticipated. Thus, we interpret
sudden deviations-from-trend in our outcome variables (transfers) in the precise period in which the event
occurs as causal effects of the event. This approach has precedent in recent literature—Dobkin et al. 2018
and Fadlon and Nielsen 2015 rely on similar assumptions in their recent event-studies of health shocks.
Using the detailed data available in the HRS, we are also able to explore potential mechanisms by which
the shocks that we consider might affect transfers, documenting the effects that each shock has on observed
household wealth, bequest intentions income, out-of-pocket medical expenses, and life expectancy.
We find that financial transfers from older households to adult children are indeed sensitive to adverse
shocks of all kinds, with wealth loss, job loss, and adverse health shocks all leading to significant reductions
in the likelihood that parents make transfers. We separately find that around the time of a widowhood,
parents’ likelihood of making transfers to their children increases, particularly when the surviving spouse is
male. Estimating the models separately for low-wealth and high-wealth households reveals that the negative
effects of adverse shocks to financial status and health on parental giving are larger both in relative and
absolute terms in households with low wealth.
Our results also reveal that adult children play an important insurance role for their aging parents. Most
notably, we find that children are much more likely to provide in-kind assistance (informal care) for their
parents following adverse shocks. Increases in informal care are especially large after the death of a spouse
and after the onset of disability or poor health, but they are present and significant for every shock that
we consider. While the overall incidence of upstream financial transfers is relatively low, we also find
that in low-wealth households, children’s likelihood of making financial transfers to their parents increases
substantially following adverse shocks to finances and health.
Taken together, our results show that adverse events in the households of aging parents have important
intergenerational effects. In particular, adult children of parents in low-wealth households receive fewer
financial transfers from their parents following negative shocks to parental wealth, employment, and health.
As previous research has shown that large financial transfers from parents to their adult children are given
in response to adverse shocks in children’s own households or as an investment in children’s human capital
(i.e., when the marginal utility from such transfers is particularly high), these reductions in large transfers
could have important welfare effects. These reductions in transfers from parents to children are paired
with increases in upstream transfers—children are also more likely to provide both money and time to their
parents following adverse shocks. The positive response of upstream transfers to parental shocks may further
exacerbate the transmission of adverse events to the younger generation. However, it also likely increases
parental welfare through added consumption smoothing.
2
Our examination of the effect of specific events on intergenerational transfers provides new insight into
the causal effects of wealth, income, family structure, and health on giving and receiving of transfers that
has not previously been gained from observational studies. Thus far, life events in giver households have
not received much attention in the large empirical literature on inter vivos transfers, which has primarily
studied the cross-sectional determinants of transfers and has focused on identifying the associations between
children’s characteristics and life events and their receipt of financial transfers (see, among others, Cox and
Rank 1992; Rosenzweig and Wolpin 1993, 1994; McGarry and Schoeni 1995; McGarry 2016). Our event-
study results suggest that the strong correlations between transfers and parental wealth, income, household
structure, and health that are seen in the cross-section are not merely reflective of spurious correlation but
rather reflect causal effects of those factors on giving.
Our results also provide new insight into the ways in which families mediate the effects of recessions.
Previous studies have focused on parents’ roles in assisting children after adverse shocks (for example, Bitler
and Hoynes 2015; Dettling and Hsu 2018; Attanasio, Meghir and Mommaerts 2018). Our study highlights
the simultaneous importance of adverse shocks in parents’ households, suggesting that they may exacerbate
the direct effects of an economic downturn on children. Our results also underscore children’s role in helping
their parents smooth consumption after shocks. Both sides of the story are relevant for understanding the
role of the family safety net in the Great Recession and, more recently, during the COVID-19 pandemic.
Finally, our findings have important implications for the value and effectiveness of social safety net
programs that might substitute for family exchange, suggesting that public income support and medical
insurance programs that aim to help households smooth their consumption following shocks to income and
health may crowd out adjustments to transfers within families. For example, the extent to which family
members increase informal care in response to health shocks affects the demand for private long-term care
insurance and the degree of potential crowd-out from a public option (Pauly, 1990; Brown and Finkelstein,
2011). On the flip side, during the spread of the COVID pandemic, caregiving relationships within multi-
household families were disrupted, with many adult children unable to provide in-person care to their aging
parents. In these unusual circumstances, the public safety net for older households may be more important
than ever (Stokes and Patterson, 2020).
2. Theoretical Models of Intergenerational TransfersWe begin our study by reviewing the fundamentals of theoretical economic models of intergenerational
transfers and considering how their predictions relate to adverse shocks in parent households. Citing the
complexity of the dynamic processes determining intergenerational transfers, economists have recently
opted not to specify and estimate structural models, choosing instead to outline key assumptions with in-
formal theoretical discussion (see, for example, Altonji and Villanueva 2007; McGarry 2016; Haider and
McGarry 2018). The canonical analytical models of inter vivos transfers (e.g., in Altonji, Hayashi and Kot-
likoff 1997) are a useful starting point, but are simplified and do not generate concrete predictions about the
effects of shocks to income, wealth, and health on transfers. In order to have a theoretical framework to aid
3
in interpretation of our empirical results, we build on the discussion in Altonji and Villanueva (2007), incor-
porating elements of recent theoretical work on long-term care insurance, late-life spending, and family risk
sharing. We also extend the discussion to include upstream transfers.
2.1. Parent-to-child transfersAltonji and Villanueva (2007) informally outline a theoretical model of the determination of parent-
to-child transfers that serves as the starting point for our discussion. In their model, parents maximize
an expected lifetime utility function that depends on their own consumption, their children’s utility, and
(optionally) their level of giving through a “warm glow” mechanism. Parents in the model start with an initial
wealth endowment and have an uncertain stream of labor earnings prior to retirement. After retirement,
they receive a flow of social security income, pension income, and labor earnings, which is a deterministic
function of their prior labor earnings and depends on their marital status. Life expectancy is uncertain, and
the flow of post-retirement income terminates when both parents are dead. In each period of the model,
parents choose how much to save from their income and wealth and how much to spend on their own
consumption and on transfers to adult children. These choices, along with the flow of labor earnings, social
security income, and pension income, determine parents’ wealth in later periods.
A key feature of this model (and most theoretical models of transfers) is the altruism motive—parents
care about their children and make transfers in order to increase their children’s utility. Under altruism, other
things being equal, increases in parental lifetime income and wealth should lead to increases in transfers.
Altruistic motives lead parents to want to transfer money earlier, when children’s own income and wealth
levels are low, the value of investments in human capital is high, and children are likely to be liquidity-
constrained (Cox, 1990). Another key ingredient of the model—uncertainty—has the opposite effect. In the
presence of uncertainty about their own future income, future consumption needs, and life expectancy, as
well as uncertainty about their children’s future incomes, parents are incentivized to delay transfers as long
as possible in order to gain more information and avoid accidentally overspending (Altonji, Hayashi and
Kotlikoff, 1997). Precautionary motives, particularly those associated with own-income risk and late-in-life
health risk, are emphasized in recent work on post-retirement consumption and the demand for long-term
care insurance (see, e.g., Ameriks et al. 2018; Laitner, Silverman and Stolyarov 2018).
Considering the effects of adverse life events in parental households on transfers in light of this model, it
is clear that the effects of a particular shock, whether it be to wealth, income, household structure, or health,
on parental giving may operate through a variety of mechanisms. First, the shock may affect parents’ ex-
pected lifetime wealth and income stream. The size of changes in lifetime wealth and income will depend on
many factors including the nature, size, and permanence of the shock, the existing accumulation of wealth,
and the availability of private and social insurance. Second, a shock may change parents’ own expected
future consumption needs. As highlighted in recent research, preferences may be state dependent. Health
status in particular, is often assumed to cause changes in the marginal utility of consumption (Ameriks et al.,
2018; Brown, Goda and McGarry, 2016; Finkelstein, Luttmer and Notowidigdo, 2013; Laitner, Silverman
and Stolyarov, 2018). Changes in perceived future consumption needs could also result from changes in
4
time use, expected health costs, household size, or life expectancy. Third, a shock may reduce the incentive
for parents to delay inter vivos giving if it reduces the level of uncertainty about future income or health.
Additionally, preferences for own consumption and for child utility, the degree of liquidity constraint, and
household discounting and risk aversion likely play important roles in mediating the effects of shocks on
transfers.
2.2. Child-to-parent transfersAbsent from Altonji and Villanueva’s model, and from many analytical models of inter vivos transfers,
is a framework for simultaneously considering upstream (child-to-parent) and downstream (parent-to-child)
transfers. In a summary article, Laferrere and Wolff (2006) outline several different models that allow
for bilateral transfers. One approach is to assume two-sided altruism. In this model, either the parent
transfers to the child (if parent income is sufficiently high or child’s income is sufficiently low), or the child
transfers to the parent (if child income is sufficiently high, or parent income is sufficiently low), or neither
makes a transfer. Under mutual altruism, children transfer money to parents who may be unable to borrow
against a fixed income stream or may encounter shocks so large that they are unable to insure against them.
Thus, adverse shocks to parents’ expected income and wealth should lead to increases in upstream transfers
(Sloan, Zhang and Wang, 2002) and the expected effects of changes in parental consumption needs and life
expectancy should mirror the effects in one-sided altruistic models.
Another approach to thinking about upstream transfers is to assume that parents make transfers in ex-
change for services or financial transfers that their children are providing or will provide to them in the future
(Cox, 1987). In an exchange model, the receipt of transfers from children might be viewed as reciprocity for
transfers given in another form (e.g., services exchanged for money), for transfers given previously, or for an
expected bequest. The exchange arrangement might be ongoing payment for informal care or an insurance
arrangement in which children receive transfers and in return provide a safety net in case of catastrophic
health expenses or unexpectedly long life. Exchange motives add another layer of complexity in consider-
ing the effects of life events, as events might not only alter parents’ ability to pay and their marginal utility
of consumption, but might separately alter the marginal utility derived from services. Moreover, there may
be a lag between payment for and receipt of services. Generally, with an exchange model, the provision
of downstream financial transfers should increase with parents’ income and wealth, and upstream financial
transfers should increase when parents’ income and wealth falls. Meanwhile, downstream transfers and the
(upstream) provision of services should both increase when the marginal utility from services increases. The
response of informal care to changes in the parents’ financial situation is difficult to predict. On one hand,
decreases in income and wealth reduce ability to pay for child services. On the other hand, if income and
wealth fall enough, recipients might be unable to afford market services, which could increase the marginal
utility derived from child-provided services.
A final alternative is to treat extended families as multilateral insurance networks, assuming that par-
ents and children help each other to smooth consumption. For example, Attanasio, Meghir and Mommaerts
(2018) apply a model of small group risk-sharing to extended families, arguing that the barriers to infor-
5
mation and risk sharing are lower within families than between households more generally. This kind of
model treats parents and children equally and would allow for transfers running in both directions at any
frequency, with transfers adjusting as the ratio of marginal utilities between parents and children fluctu-
ates. Like the other models described above, this model also generates the prediction that adverse shocks to
parental household circumstances should lead to increases in upstream transfers.
Taken together, these models provide a useful framework for examining the effect of adverse shocks to
parental households on transfers from children to their parents. First, all three models imply that reductions
in parents’ future wealth or income may lead to more transfers from their children. Second, they predict that
parents with lower wealth at the time of the event will be more likely to receive transfers. This effect may
be due to the parents’ inability to absorb large shocks themselves, as in risk-sharing or 2-sided altruism,
or due to a rising marginal utility of child provided services, as in exchange models. Third, the models’
implications for the effects of health shocks on transfers from children to parents are ambiguous. On the one
hand, health shocks could lead to lower expected resources in the future, due to higher medical expenses or
lower income, increasing child to parent giving. On the other hand, if the health shock substantially resolves
uncertainty around future income and expenses, then they may need less help. Finally, the models do not
provide guidance on when transfers from children to their parents will be financial or in the form of services.
The mixture of support is an important empirical question as different forms of transfers may vary in their
effectiveness and cost.
3. DataIn light of the many theoretical mechanisms listed in the previous section, and as discussed in the ap-
pendix, both the qualitative and quantitative effects of adverse shocks in parental households on transfers
are difficult to predict from theoretical models alone. Thus, we turn to empirical analysis in order to better
understand these effects. In this section, we begin with a description of the data that we use for our analy-
sis. In the following section, we conduct descriptive analyses of downstream transfers with an emphasis on
parental circumstances and an expanded focus that includes upstream transfers. Then, for our main analysis
we use event-study methods that highlight discrete changes in transfers that occur around the time of each
specific adverse event.
3.1. Data SourceFor our empirical analysis, we use eleven survey waves spanning 1993 to 2014 from the Health and
Retirement Study (HRS) RAND public-use data files.4 The HRS is a panel survey that comprises a series of
national probability samples of Americans over the age of 50. To keep the panel representative of its target
population, the HRS has recruited cohorts of participants every few years. To date, there are six cohorts,
4The HRS (Health and Retirement Study) is sponsored by the National Institute on Aging (grant number NIA U01AG009740)and is conducted by the University of Michigan. The data files we use were produced by RAND with funding from the NationalInstitute on Aging and the Social Security Administration.
6
born between 1921 and 1959. We combine data from the core HRS survey, which focuses on the main
respondents, and the family survey, which provides detailed information about respondents’ children and
transfers between family members. Our data set includes households with members between 50 and 85
years old with at least one child over the age of 18 (we exclude children under 18 from our analysis). In
order to minimize within-household changes in reporting, we use the reports of household and individual-
level variables from the longest-lived respondent. We adjust all measures that are reported in nominal dollars
to 2017 dollars using the Consumer Price Index. Throughout our analysis, we weight observations using
HRS household weights. In our main estimation sample, we have 19,384 unique households and 113,781
household-wave observations.
3.2. Intergenerational TransfersThe HRS gathers detailed information on the health, employment, and finances of respondents and their
children. Importantly, we observe large financial transfers that each child received from, or gave to, their
parents in each wave of the survey. Specifically, the survey asks respondents whether they had provided
financial assistance totaling at least $500 to any child since the last interview. Financial assistance is defined
as giving money directly, helping to pay bills, or covering specific costs (e.g., nursing home care, rent) and
can be considered support, a gift, or a loan.5 We focus on the extensive margin of giving for our analysis.
In particular, we create an indicator for whether the household made a large transfer to at least one child
during the survey wave and another indicator for whether the household received a large transfer from at
least one adult child.6 Starting in Wave 3, the HRS also collects data on whether children help their parents
with activities of daily living (ADLs) and instrumental activities of daily living (IADLs). ADLs are basic
self-care tasks such as walking, dressing, or bathing. IADLs are more advanced self-care tasks that include
house cleaning, shopping, and managing transportation. HRS respondents also report whether their children
help them with any money management tasks, such as balancing a checkbook or filing taxes.
3.3. Other VariablesWe use a comprehensive household income measure that includes all income for the respondent and
his or her spouse during the past calendar year. This measure includes regular earnings as well as unem-
ployment and workers’ compensation, Social Security and other retirement income, interest and dividends,
alimony/child support, capital income (e.g., businesses, gross rent), and all cash assistance from outside the
household. It does not include capital gains or in-kind transfers (e.g., SNAP benefits). Our primary measure
of household wealth is the net value of all household assets asked about by the HRS. These include the
reported value of the household’s primary residence, nonresidential real estate, all transportation vehicles,
5In the HRS, financial assistance excludes shared housing and food as well as inheritances.6Transfer amounts are also reported, but they are noisy and often imputed in the data. Additionally, given the long recall period
and the lack of verification, it is likely that even reported transfer amounts contain substantially more measurement error thanwhether any transfers were made.
7
the value of any businesses, IRA and Keough accounts, non-IRA stock holdings, checking accounts, certifi-
cates of deposit, bonds, and other reported assets.7 For all income and wealth variables, we use RAND’s
imputations when necessary.
In our descriptive analysis, we use information on the age and race (white/nonwhite) of the HRS respon-
dent and his/her spouse, household structure (couple, single male, single female), an indicator for respondent
or spouse being disabled, an indicator for either respondent or spouse being in poor health, an indicator for
hospitalization during the survey wave, employment status, home ownership, supplemental security income
(SSI) receipt, and number and proximity of adult children.
To aid in interpretation of our main results, we use information on life expectancy, expected bequests,
and out-of-pocket medical expenses. For life expectancy, the HRS asks respondents to report the probability
that they will live until 75, 85, or for 10 more years, depending on the age of the respondent. We combine
these three variables based on data availability.8 For expected bequests, respondents and their spouses are
asked to report the probability that they will leave a bequest of at least $100,000. Finally, the HRS collects
detailed information on medical expenditures in the last two years or since the last interview. Medical
expenditures cover a broad range of payments, including hospital and nursing home costs, doctor, dentist
and outpatient surgery costs, prescription drug costs, and home health care and special facilities or services
costs.
3.4. Defining Adverse EventsWe define life events by identifying changes between waves, which is similar to the approach used in
Dobkin et al. (2018) and McGarry (2016). We define a negative wealth shock as the loss of at least 25
percent of household net wealth from the previous wave. In order to avoid including changes that represent
very small dollar values, we construct this shock only for households with at least $20,000 of total reported
wealth. Forty-four percent of HRS households in our sample experience an adverse wealth shock under
this definition.9 We code a respondent as experiencing a job exit if they report working for pay last wave
and report not working for pay in the current wave. This definition excludes any unemployment spells that
are shorter than the time between waves and includes transitions out of the labor force due to retirement
or a shift to volunteer work. Forty-one percent of households in our sample experience a job exit by this
definition.
A respondent experiences widowhood if they are married in the prior wave and report that their partner
7The wealth measure is net of any reported debt and does not include secondary residences.8To elicit these probabilities, the respondent is asked “(What is the percent chance) that you will live to be (age XX) or more?”
Respondents are asked to report probabilities as a number between 0 and 100, with 0 indicating “absolutely no chance” and 100indicating “absolutely certain.” These questions seem to elicit valid probabilities, as their averages are similar to average lifeexpectancy at the population level and are correlated with variables that affect life expectancy, such as socioeconomic status, in thesame way that actual survival probabilities do (Hurd and McGarry, 1995).
9The wealth measure is somewhat noisy, bouncing around from wave to wave within households. In order to avoid picking upwealth shocks that resulted from measurement error (unusual positive values), we ignore negative wealth shocks for which therewas a positive wealth shock of at least 50 percent in either of the two previous waves.
8
is no longer alive in the current wave. This occurs at some point in 15 percent of households in our sample.
Following Fadlon and Nielsen (2015), we separately examine male widowhood (female death) and female
widowhood (male death). We additionally examine three morbidity shocks. First, we use an indicator for
whether the respondent or spouse was hospitalized during the survey wave. Hospitalization has previously
been found to be associated with substantial increases in health expenditures, reductions in labor supply,
and increases in household debt (Dobkin et al., 2018). This is the most common health shock in our data,
occurring at some point in 70 percent of HRS households. We also create an indicator for the onset of
disability, given its importance to future earning potential and expected future medical costs. We define
disability onset to be the first wave in which the respondent or spouse reports difficulty with any ADL or
IADL. By separating out disability onset, we can pick up changes in patient health that influence future
income and medical expenses but may not be perceived as a large decline in general self-reported health.
Disability onset occurs at some point in 51 percent of households. We additionally use self-reported health
to capture a broad range of severe and sudden health shocks. In each wave, both the respondent and spouse
are asked to report their general health status on a scale from 1 (Excellent) to 5 (Poor). We define a health
shock as a wave in which respondents who previously reported being in good health or better (a 3 or lower)
report being in poor health (a 5). This shock occurs at least once in ten percent of HRS households. In
order to avoid the influence of households who are experiencing declining health trends, we focus on health
shocks that are occurring for the first time following two waves without a similar shock.
4. Correlates of Intergenerational Transfers in Older HouseholdsWe begin our empirical analysis with a set of descriptive findings. These results serve three purposes
in our paper. First, they establish the baseline propensities to give and receive in HRS households and the
frequencies of upstream and downstream transfers, providing important insight into the nature of intergener-
ational transfers in older households. Second, they reveal cross-sectional heterogeneity across households in
the propensities to give and receive along dimensions that are associated with the shocks that we consider—
wealth, income, household structure, and health. Finally, panel fixed effects regressions provide preliminary
evidence that those variables are causally associated with the giving and consistent with the predictions of
the theoretical models presented in Section 2.
Table 1 shows summary statistics for our full sample and for household waves in which parents make or
receive transfers. We observe large financial transfers to adult children in around 35 percent of household-
wave observations. Households that transfer money to their children are younger and more educated than
the average HRS household. Giving households are also in better health, have higher per-capita income and
assets, and report a higher probability of leaving a bequest for their children. Upstream transfers are less
common; we observe transfers from children to parents in only six percent of household-wave observations.
Compared with the full sample, households receiving financial transfers from their children are older, more
likely to have a non-white head or spouse, and have substantially lower income and wealth than the average
HRS household. The rate of receipt of help from children is also relatively low in our sample, and we find
9
that recipients of help are even more disadvantaged, particularly in terms of health.10
In Table 2, we examine rates of giving and receiving, along with average transfer amounts separately
by age, wealth percentile, employment status, household structure, reported health, and disability status.
Among households that transfer money to their children, the average amount given across an HRS wave is
around $13,561. Giving is more common among younger households—45 percent of households between
the ages of 50 and 60 made large transfers to their children, while only 27 percent of households between
75 and 85 did. Not surprisingly, giving is also far more common in wealthy families, with almost half of
all households in the top wealth quintile reporting large transfers. Meanwhile, rates of giving are lower in
households with at least one member disabled or in poor health and in single female households.
Rates of receipt of transfers also vary substantially across groups, with upstream financial transfers
most common in the lowest wealth quintile, households with poor health or disability and single female
households. Among recipient households, the average amount received from children during a wave is just
over $5000. Receipt of ADL/IADL assistance also varies widely, particularly along the health dimension,
ranging from 1 percent in households with excellent health to a full 24 percent in households who report
having poor health. Notably, single females, households in the bottom wealth quintile, and households in
the oldest age group, receive substantially more help from their children than the average household, at a
rate of 11 or 12 percent. These statistics suggest that, despite their relatively low rate of overall occurrence,
child-to-parent transfers may play an important safety-net role.
While summary statistics provide suggestive evidence that household wealth, income, structure, and
health status are important determinants of the giving and receiving of intergenerational transfers, it is diffi-
cult to discern the relative importance of each variable as a predictor of transfer behavior from sample means
because the variables are correlated with one another. Cross-sectional and fixed-effects regressions, which
have been widely used in the literature on inter vivos transfers (for example, in Cox 1990 and McGarry
2016), allow us to determine whether variation in each factor is associated with variation in transfers, even
after the other factors (and even time-invariant family characteristics) are accounted for.
A few notable patterns emerge from our regression results, which are presented in Table 3. First, parental
giving remains positively correlated with household wealth and income, both across and within households,
even when controlling for a wide range of other household characteristics and household fixed-effects. The
probability of receiving a financial transfer is negatively correlated with income and the probability of re-
ceiving help from your children is negatively correlated with wealth. This pattern suggests that baseline
levels of wealth and income will likely influence both the degree to which transfers respond to adverse
shocks and the form that these transfers will take.
Second, giving does not appear to be associated with household health status or disability once other
variables have been included in the model. By contrast, the receipt of both transfers and help in older
10The high incidence of reported disability is mechanical, in some sense, because respondents must report that they “needassistance” with an activity in order to report receiving any help. However, respondents who need assistance do not necessarilyreceive assistance, and assistance does not necessarily need to come from children.
10
households is clearly associated with disability and poor health. Models of parental giving suggest that
the small association between parental transfers and health may be due to the offsetting effects of health
shocks. Variation in health may both reduce giving by directly impacting parental resources and increase it
by resolving uncertainty around life expectancy.
Finally, both giving and receiving transfers is correlated with household structure, even in household
fixed-effect models. Relative to two-person households single-male households are more likely to give fi-
nancial transfers and receive help and single-female households are more likely to receive financial transfers
and help. These patterns are consistent with an exchange model, which suggests that a transition from a
two-person household to a single-person household increases the marginal utility of services from children.
This increase leads to children being more likely to help with tasks and parents more likely to give transfers
to their children in return.
5. Effects of Adverse Life Events in Parental HouseholdsIn order to gain insight into the dynamics of intergenerational transfers within households and how trans-
fers respond to changes in household circumstances, we conduct a series of event-studies for a set of adverse
events affecting wealth, income, household structure, and health. Thus far, event-study analysis has not been
used to study intergenerational transfers. One reason for this is that household events such as job loss and
health shocks cannot be considered purely exogenous, and thus are unlikely to satisfy a “no pre-trends”
assumption. Here, we show that careful examination of changes in intergenerational transfers surrounding
household events reveals important information about the dynamics of transfers within households that can-
not be obtained from fixed effects models, which confound slow evolutionary changes in outcomes with
discrete changes.
5.1. Event-Study ResultsWe begin by considering our two financial shocks—wealth loss and job exit. To provide support for
our “shocks” identification strategy, we first examine time patterns in intergenerational transfers during a
period in which a large number of households experienced these two types of shocks—the Great Recession.
In Figure 1, we examine the raw changes in both downstream and upstream financial transfers surrounding
each shock. In particular, we focus on 1,376 cases of wealth loss and 1,083 cases of employment exit
reported during wave 10 of the survey, which was conducted in 2010 and referred to the 2008-2010 period,
and examine the trajectories of giving across HRS waves 7 through 12 (relative to wave 8 levels) for those
households compared with households who were present for the Wave 10 survey but did not experience
a shock at that time. This is essentially a simple difference in differences approach. We find that the
trajectories in both giving and receipt of transfers were almost indistinguishable between the two groups
leading up to wave 10 and then diverge. Figure 1 reveals that households that did not experiences wealth
and employment shocks during the Great Recession were actually more likely to make transfers to their
adult children during wave 10, while the rate of giving for households that experienced shocks decreases
11
starting in wave 10. The figure also shows small increases in child-to-parent transfers in the wave of the
shock, particularly following wealth loss.
For our full graphical event-study results, we estimate the following model for events occurring across
all HRS waves:
Tht = αh + γt +
3+∑s=−2
1(t = t∗h + s)βs +Aψ + εht (1)
where Tht is a dummy for whether household h made (or received) a transfer at time (wave) t, αh is a
household fixed effect, γt is a wave fixed effect, A is (quadratic) age, t∗h is the event time for household h,
and s is the number of waves since the event. Recall that HRS waves are generally two years apart, which
means that the event in question occurs sometime in the two years prior to the period-zero interview, the
period-one interview occurs 2–4 years later, the period-two interview occurs 4–6 years later, and so on. We
omit indicators for three or more waves prior to treatment so that β represents average outcomes in each
wave relative to the average for the period more than six years prior to treatment. We assume that any
dynamic effects of treatment are no longer relevant after six years and combine three or more waves after
the event into a single indicator.
It is likely the case that families which experience a given event are different in important and unob-
servable ways from families that do not. Throughout our event-study analysis, we restrict our sample only
to households who ever experience a given event to avoid confounding the effect of an event with the ef-
fects of these differences. In this approach, the effects of an adverse event are identified if their timing is
quasi-random. In particular, households that experience the event must be on similar trends before the event
occurs, and the event must be unanticipated. The specification in Equation 1 provides a natural test of this
assumption. If any of the coefficients on the pre-waves are different from 0, then the event timing is likely
to be endogenous or families can anticipate the event and adjust.
Following Dobkin et al. (2018), we also estimate a parametric event-study in which we restrict pre-event
effects to follow a linear trend and flexibly estimate post-treatment deviations from that trend. This allows
us to summarize the event-study estimates in fewer coefficients, which is helpful for examining mechanisms
and conducting heterogeneity analysis, and allows us to estimate deviations from any pre-event trends that
may exist. In our results tables, we focus on periods 0 and/or 1 after each shock, which represent the
two-year period in which the shock occurs and the period 2–4 years following the shock, respectively.
Tht = αh + γt + (t− t∗h)φ+9∑
s=0
1(t = t∗h + s)βs + εht (2)
Figure 2 shows event-study estimates, generated with Equation 1, of the effects of shocks to parental
wealth and employment on upstream and downstream financial transfers. Coefficients on β1 from estimating
Equation 2 are presented in the first two columns of Table 4. Recall that in Section 2.1 we predicted that the
loss of wealth and earned income would lead to decreases in downstream transfers and increases in upstream
transfers. In fact, we do find that both wealth loss (in Panel A) and job exit (in Panel B) significantly reduce
the likelihood that parents make large financial transfers to their children. In Table 4, we find that two
12
waves after wealth loss, the likelihood of giving a transfer is reduced by more than 5 percentage points.
This represents a twelve percent reduction in the likelihood of transfer following wealth loss relative to the
pre-event mean for that estimation sample. The estimated treatment effect from Equation 2 is −0.036, a
nine percent reduction during the period 2-4 years after the shock. For job exit, we find a slightly smaller,
3.5 percentage point (eight percent) reduction in the likelihood of giving.
Looking at upstream financial transfers, we see evidence in Figure 2 of small increases in the likeli-
hood of receiving a large financial transfer from a child following wealth loss and job exit. The regression
coeffients in Table 4, which account for pre-shock trends, show an increase of one percentage point (25 per-
cent) following a negative wealth shock but are not statistically significant for employment exit. Together,
the results in Table 4 and Figures 1 and 2 suggest that events that alter the financial situation in aging house-
holds directly impact the adult children of those households, reducing the likelihood that they receive large
financial transfers from their parents and causing them to increase financial transfers to their parents.
Next, we repeat our event-study exercise, now estimating the effects of fatal and non-fatal health shocks.
Figure 3 shows event-study estimates of the effects on intergenerational transfers of widowhood, while
Figure 4 shows the effects of three morbidity shocks— 1) hospitalization, 2) disability onset, and 3) the
transition to “poor” self reported health. When considering health shocks, we additionally show graphical
evidence of the effects of these shocks on the likelihood that children provide in-kind assistance (help) to
their parents, presented in the third column of graphs.
Because the regression results in Table 3 suggest that single male households and single female house-
holds have different patterns of giving and receiving, and because Fadlon and Nielsen (2015) demonstrate
differing effects of fatal shocks to males and females, we separately examine male and female widowhood.
In Figure 3, we find that the patterns are similar between the two panels, but there is a much larger increase
in giving following male widowhood (female death) and a more-sustained increase in helping following
female widowhood (male death). These results are echoed in the regression results in Table 4, which show
that male widowhood is associated with a full 12.6 percentage-point increase in the likelihood of a large
financial transfer in the period of the event—a 37 percent increase—and an 8.1 percentage-point increase in
the period after the event. Turning to upstream transfers, we see significant increases of around 2 percent-
age points (increases of more than 50 percent) in upstream financial transfers during the event periods for
both male and female widowhood, which are not sustained in the next period. We also see very large and
sustained increases in in-kind assistance (helping) after widowhood for both sexes.
Event-study estimates of the effects of morbidity shocks are shown in Figure 4. Across the four panels,
we do not see clear evidence that health shocks cause reductions in parent-to-child transfers. While there
appear to be some reductions in giving in each post-event period, it is difficult to discern the magnitude of any
drop in giving that can be attributed to the event visually because the pre-trends are not flat,. The parametric
estimation results in Table 4 show a statistically significant trend-deviation of around 2 percentage points (5
percent) in both the period of hospitalization and in the 2-4 years following hospitalization. However, there
are no significant changes in giving following disability onset or poor health.
Looking at upstream transfers, we find the opposite pattern–hospitalization has no effect on upstream
13
financial transfers and very little effect on helping, while both disability onset and the onset of poor health
lead to statistically significant increases in child-to-parent financial transfers and to very large increases in
the likelihood that children provide assistance with ADLs. In fact, the onset of poor health is associated
with a 37 percent increase in the probability of receiving financial assistance from one’s children (a 1.7
percentage point increase) and a full 231 percent increase in the likelihood of receiving in-kind assistance
(8.1 percentage points). These findings suggest that children’s aid to their parents is closely tied with their
parents’ health status, and is consistent with the exchange and informal insurance models discussed in
Section 2.2.
Taken together, the results in Figures 2, 3, and 4 and in Table 4 show that parental giving and receipt
of transfers is sensitive to adverse shocks to household circumstances. We see reductions in parental giving
following adverse shocks to household wealth and income, including wealth loss, job exit, and hospitaliza-
tion, and we see large increases in giving following the death of a parent, particularly if the surviving parent
is male. Meanwhile, adult children increase financial transfers to their parents in response to all kinds of
adverse shocks, but those increases in transfers are transitory, occurring only in the periods in which the
shocks occur. Finally, we see that children increases their provision of help to their parents following every
shock that we consider, with the largest increases occurring in response to widowhood, disability onset, and
poor health.
Thus far, our analysis has intentionally focused on the effects of non-specific health events—death,
hospitalization, disability onset, and self-reported “poor” health—on intergenerational transfers. This focus
increases the precision of the analysis since a larger number of households experience the more general
events. In Table 5, we present the effects of a new onset of four specific health conditions—cardiac events,
stroke, memory conditions, and cancer—on intergenerational exchange as a check for whether the less
specific health events are not covering up significant heterogeneity across conditions. Across these four
events, the results are qualitatively similar to the other health shocks we consider in the rest of the paper:
health shocks have small effects on parental giving but they induce financial transfers and help from children.
However, there are two main differences. First, cancer diagnosis results in much larger reductions in parental
giving to adult children. Cancer diagnosis reduces the likelihood of giving by 3.8 percentage points, or 9.6
percent of the baseline mean, in the 2–4 years after diagnosis. Second, memory conditions and cancer
diagnosis lead to significant increases in child-to-parent financial transfers but cardiac events and stroke do
not.
5.2. Possible First Stage MechanismsWe examine the mechanisms underlying the effect of adverse shocks on transfers by estimating their
“first-stage” effects on mechanisms suggested by the theoretical models presented in Section 2. We gen-
erate these results by estimating the event-study model in Equation 2 with summary measures of the key
theoretical mechanisms on the left-hand side. In particular, we explore mechanisms related to current and
future parental resources using detailed measures of household per-capita wealth and income. We also ex-
amine the effects of adverse events on out-of-pocket medical spending which often spikes in the wave of
14
the event and may lead to liquidity issues for respondents. Additionally, adverse events, particularly health
shocks, may influence current giving by altering expectations about future needs. We capture this path-
way by looking at effects on self-reported probability of leaving a bequest, and self-reported probability of
surviving past the next major age milestone (75, 85, or 10 years, depending on respondent’s age).
We generate these results by estimating Equation 2 with a set of household characteristics on the left-
hand side: real household per-capita wealth, bequest plans, real household per-capita income, household out-
of-pocket medical expenses, and a variable summarizing life expectancy based on the respondent’s reported
probability of surviving past the next major milestone (75, 85, or 10 years, depending on respondent age).
These results, presented in Table 6 provide us with important insight into the nature of the events that we
consider and how they alter households’ perceived financial status and planning window.
The wealth shocks that we consider are associated with large reductions in per-capita wealth (by con-
struction) and also with around a five percentage-point reduction in the expected probability of leaving a
bequest of more than $100,000. Perhaps surprisingly, Table 6 reveals that wealth shocks are not associated
with significant reductions in household income per capita. By contrast, job exit is associated with large
reductions in household income and with small increases in out-of-pocket medical expenditures (possibly
from loss of employer-sponsored insurance), but is not associated with significant changes in wealth or
bequest plans.
Turning to health shocks, Table 6 confirm that widowhood is associated with increases in both per capita
wealth and per capita income. Male widowhood, but not female widowhood, is also associated with an
increase in the reported likelihood of leaving a bequest and both are associated with large reductions in
household medical expenses (likely reflecting high medical costs prior to death for the spouse who passed
away). None of the three health shocks are associated with changes in the likelihood of leaving a bequest,
but all three are associated with increases in out-of-pocket health expenditures and reductions in life ex-
pectancy. Hospitalization is associated with the largest increase in out-of-pocket health expenditures and is
also associated with reductions in reported wealth and earned income, which is consistent with the findings
of Dobkin et al. (2018).
The results in Table 6 support our interpretation of several adverse events—wealth shocks, job exit,
widowhood, and hospitalization—as significant financial shocks. Wealth loss and male widowhood in par-
ticular lead not only to changes in household per-capita wealth and income, but also changes in households’
bequest plans, which suggests they have altered households’ perceptions of future financial security, while
job exit and hospitalization lead to reductions in income and increases in health expenditures. The other two
morbidity shocks—disability and poor health onset—do not have large immediate effects on households’
broader financial situations (wealth and income) and instead are associated only with small (in absolute
terms) increases in out-of-pocket medical expenditures and reductions in life expectancy.
5.3. Heterogeneity by WealthThe results from our event-study analysis suggest that parental wealth and income are closely tied with
the level of parental giving. There is also evidence that parental need (disability and poor health status in
15
particular) drives parental receipt of transfers and time from children. An important open question is whether
the full-sample effects are masking differing effects that depend on the level of accumulated household
wealth. There are two reasons for considering wealth heterogeneity. First, recall from Section 2.2 that it is
natural to think of high-wealth parent households as being givers of transfers (or at the margin of giving) to
their children and lower-wealth parent households as being recipients of transfers from their children. Thus,
it is possible that only some households are in a range of either giving or receiving money. Second, wealthier
households may be less liquidity-constrained, and thus are better able to smooth their own consumption (and
giving) across periods even in the presence of shocks.
In order to examine wealth heterogeneity, we average per-capita wealth across waves for each household
and stratify our sample by average real per-capita wealth, separately examining the effects of events on
giving for households above and below the median wealth level. These results are presented in Table 7.
In order to aid interpretation, we show effects on first stage mechanisms separately by wealth in Appendix
Tables A2 and A3.
Consistent with the theoretical models, we find that the effects of adverse household events on intergen-
erational financial transfers are concentrated in low-wealth households, and most of the estimated effects
are larger, both in absolute and relative terms, than those from the full sample. In the low-wealth sample,
the estimated effect of wealth loss on the probability of giving a transfer is -6.1 percentage points, or 18.7
percent of the baseline mean for that sample. The estimated reduction in the likelihood of giving after job
exit is 12.7 percent of the baseline mean, and the increase associated with male widowhood is 39 percent.
In this group, hospitalization is found to reduce giving in period of the event and the onset of poor health
is associated with reduced giving in the following period. In the high-wealth sample, we still see substan-
tial increases in giving following parental death, and male widowhood in particular (an increase of 14.8
percentage points, or 37 percent), and also see smaller reductions following hospitalization and job exit.
Considering upstream transfers, stratifying by wealth reveals large and significant increases in child-
to-parent transfers following wealth loss, job exit, male and female widowhood, disability onset, and poor
health in low-wealth families. Following a negative wealth shock, the likelihood of receiving a large financial
transfer increases by 2.1 percentage points, or 33.9 percent. Parental death, disability onset, and poor health
lead to even larger increases in transfer receipt, with the largest increase of 4.4 percentage points (70 percent)
found in response to female widowhood. We find no evidence at all of changes in upstream financial transfers
in high-wealth households. We do, however, find large adjustments in children’s provision of assistance to
their parents in both high- and low-wealth households following adverse events, with low-wealth households
showing larger absolute adjustments and high-wealth households showing larger adjustments relative to their
(lower) baseline mean rates of receipt of help.
6. Discussion and ConclusionIntergenerational transfers are an important economic phenomenon. Gale and Scholz (1994) estimate
that intended inter vivos transfers account for at least 20 percent of U.S. wealth. In the Health and Retire-
16
ment Study, 62 percent of households make a large financial transfer to their adult children at some point,
and 34 percent receive either financial or in-kind assistance from their children. These transfers between
family members potentially play an important role in intertemporal consumption smoothing within house-
holds. However, research has only just begun to explore the dynamics of transfers and how they respond
to changes in circumstances within households. In this study, we use panel data to examine the association
between parental household circumstances and the likelihood that parents give money to or receive money
or help from their children. In addition to standard descriptive statistics and regression analyses, we con-
duct an event-study analysis, examining change in transfers following adverse events in parent households.
Our analysis adds to the new literature on the dynamics of transfer behavior within households (McGarry,
2016; Haider and McGarry, 2018) and also reveals potentially important intergenerational effects of adverse
events.
Our results reveal that in households with low wealth holdings, the likelihood that parents make large fi-
nancial transfers to their adult children is very sensitive to household wealth and income, falling substantially
after wealth loss and job exit and increasing after widowhood. Parental giving in low-wealth households
also falls after a hospitalization and the onset of poor health. Taken together, these results suggest that
the adult children of low-wealth parents are likely to experience negative effects of adverse financial and
health events in their parents’ households. The implication of this is that any direct effects of aggregate eco-
nomic downturns, such as the Great Recession and the recession associated with the COVID-19 pandemic,
that children experience may be enhanced by the negative spillover effects from events in their parents’
households. Given that children are known to receive transfers during times when their marginal utility
of consumption is high (McGarry, 2016), the welfare costs of the reductions in transfers that we estimate
may be large. In high-wealth households, we find some significant adjustments in giving following adverse
shocks, though those adjustments are smaller in both absolute and relative terms, and we do not find that
health shocks alter giving behavior.
Considering upstream transfers, we find that low-wealth households receive financial assistance from
their children following a wide range of adverse shocks, though these increases in upstream transfers appear
to be transitory. Importantly, we find that children also respond to adverse events in their parents’ households
by substantially increasing their in-kind assistance to their parents—we see sudden and large increases in
the likelihood that children help with activities of daily living following adverse events, particularly after
widowhood and adverse health shocks. Given that the timing of adverse shocks in parental households may
be unpredictable, parents’ sudden increased need for help following adverse shocks also place an additional
welfare burden on the children who provide them with in-kind assistance. The increase in assistance that we
observe is consistent with exchange and insurance models of intergenerational transfers—these upstream
in-kind transfers may be given in exchange for financial transfers that were given previously. Increases in
family assistance following adverse shocks have important implications for our understanding of demand
for long-term care insurance, and for potential crowdout from a publicly provided long-term-care option. To
our knowledge, we are the first to document these sharp increases in helping following adverse events using
an event-study framework.
17
The findings in this paper are especially relevant in the post-COVID era, as the pandemic has broadly
generated adverse shocks to wealth, employment, and health. Our results suggest that adult children of low
wealth households may experience a double-whammy effect of the pandemic, as any wealth, employment,
and health shocks experienced by their parents are transmitted to them in the form of reduced financial assis-
tance during a time in which they too may be experiencing wealth loss, job loss, and health hazards. Mean-
while, the capacity that children have to help to cushion shocks for their parents, particularly with increases
in in-kind assistance, may be limited during the pandemic as interpersonal contact is drastically limited
due to the risk of disease transmission (Stokes and Patterson, 2020). Thus, policymakers will need to take
into account that older households may be left without the same informal safety net that they traditionally
have following shocks to health and finances. Further research on changes in the nature of intergenerational
exchange during the pandemic era will be important when data become available.
18
ReferencesAltonji, Joseph G., and Ernesto Villanueva. 2007. “The Marginal Propensity to Spend on Adult Children.”
The BE Journal of Economic Analysis & Policy, 7(1): 1–52.
Altonji, Joseph G., Fumio Hayashi, and Laurence J. Kotlikoff. 1997. “Parental Altruism and Inter VivosTransfers.” Journal of Political Economy, 105(6): 1121–1166.
Ameriks, John, Joseph Briggs, Andrew Caplin, Matthew D. Shapiro, and Christopher Tonetti. 2018.“The Long-Term-Care Insurance Puzzle: Modeling and Measurement.” National Bureau of EconomicResearch Working Paper 22726, Cambridge, MA.
Attanasio, Orazio, Costas Meghir, and Corina Mommaerts. 2018. “Insurance in Extended Family Net-works.” Institute for Fiscal Studies Working Paper W18/17, London, UK.
Bitler, Marianne, and Hilary Hoynes. 2015. “Living Arrangements, Doubling Up, and the Great Reces-sion: Was This Time Different?” American Economic Review, 105(5): 166–170.
Brown, Jeffrey R, and Amy Finkelstein. 2011. “Insuring Long-Term Care in the United States.” Journalof Economic Perspectives, 25(4): 119–142.
Brown, Jeffrey R., Gopi Shah Goda, and Kathleen McGarry. 2016. “Heterogeneity in State-DependentUtility: Evidence from Strategic Surveys.” Economic Inquiry, 54(2): 847–861.
Chan, Sewin, and Ann Huff Stevens. 2001. “Job Loss and Employment Patterns of Older Workers.” Jour-nal of Labor Economics, 19(2): 484–521.
Christelis, Dimitris, Dimitris Georgarakos, and Tullio Jappelli. 2015. “Wealth Shocks, Unemploy-ment Shocks, and Consumption in the Wake of the Great Recession.” Journal of Monetary Economics,72(C): 21–41.
Cox, Donald. 1987. “Motives for Private Income Transfers.” Journal of Political Economy, 95(3): 508–546.
Cox, Donald. 1990. “Intergenerational Transfers and Liquidity Constraints.” Quarterly Journal of Eco-nomics, 105(1): 187–217.
Cox, Donald, and Mark R. Rank. 1992. “Inter-Vivos Transfers and Intergenerational Exchange.” Reviewof Economics and Statistics, 74(2): 305–314.
Dettling, Lisa J., and Joanne W. Hsu. 2018. “Returning to the Nest: Debt and Parental Co-Residenceamong Young Adults.” Labour Economics, 54(October): 225–236.
Dobkin, Carlos, Amy Finkelstein, Raymond Kluender, and Matthew J. Notowidigdo. 2018. “The Eco-nomic Consequences of Hospital Admissions.” American Economic Review, 108(2): 308–352.
Edwards, Kathryn. 2019. “Who Helps the Unemployed? Workers’ Receipt of Private and Public Trans-fers.” RAND Education and Labor Working Paper WR-1315, Santa Monica, CA.
Fadlon, Itzik, and Torben Heien Nielsen. 2015. “Family labor supply responses to severe health shocks.”National Bureau of Economic Research.
Fadlon, Itzik, and Torben Heien Nielsen. 2019. “Family health behaviors.” American Economic Review,109(9): 3162–91.
19
Farber, Henry S. 2017. “Employment, Hours, and Earnings Consequences of Job Loss: U.S. Evidencefrom the Displaced Workers Survey.” Journal of Labor Economics, 35(S1): S235–S272.
Finkelstein, Amy, Erzo F. P. Luttmer, and Matthew J. Notowidigdo. 2013. “What Good is Wealth withoutHealth? The Effect of Health on the Marginal Utility of Consumption.” Journal of the European EconomicAssociation, 11(S1): 221–258.
Gale, William G., and John Karl Scholz. 1994. “Intergenerational Transfers and the Accumulation ofWealth.” Journal of Economic Perspectives, 8(4): 145–160.
Goda, Gopi Shah, John B. Shoven, and Sita Nataraj Slavov. 2013. “Does Widowhood Explain GenderDifferences in Out-of-Pocket Medical Spending among the Elderly?” Journal of Health Economics,32(3): 647–658.
Haider, Steven J., and Kathleen McGarry. 2018. “Parental Investments in College and Later Cash Trans-fers.” Demography, 55: 1705–1725.
Health and Retirement Study. n.d.. RAND HRS Public Use Datasets. Ann Arbor, MI. Produced and dis-tributed by the University of Michigan with funding from the National Institute on Aging (Grant No. NIAU01AG009740). Ann Arbor, MI.
Hoynes, Hilary, Douglas L. Miller, and Jessamyn Schaller. 2012. “Who Suffers during Recessions?”Journal of Economic Perspectives, 26(3): 27–48.
Hurd, Michael D., and Kathleen McGarry. 1995. “Evaluation of the Subjective Probabilities of Survivalin the Health and Retirement Study.” Journal of Human Resources, 30(1995): S268–S292.
Kaplan, Greg. 2012. “Moving Back Home: Insurance against Labor Market Risk.” Journal of PoliticalEconomy, 120(3): 446–512.
Laferrere, Anne, and Francois-Charles Wolff. 2006. “Microeconomic Models of Family Transfers.” InHandbook of the Economics of Giving, Altruism, and Reciprocity. Vol. 2, , ed. S. Kolm and Jean MercierYthier, 889–969. Amsterdam:Elsevier.
Laitner, John, Dan Silverman, and Dmitriy Stolyarov. 2018. “The Role of Annuitized Wealth in Post-Retirement Behavior.” American Economic Journal: Macroeconomics, 10(3): 71–117.
McGarry, Kathleen. 2016. “Dynamic Aspects of Family Transfers.” Journal of Public Economics,137(C): 1–13.
McGarry, Kathleen, and Robert F. Schoeni. 1995. “Transfer Behavior in the Health and RetirementStudy: Measurement and the Redistribution of Resources within the Family.” Journal of Human Re-sources, 30(1995): S184–S226.
Munnell, Alicia H., and Matthew S. Rutledge. 2013. “The Effects of the Great Recession on the Re-tirement Security of Older Workers.” Annals of the American Academy of Political and Social Science,650(1): 124–142.
Pauly, Mark V. 1990. “The Rational Nonpurchase of Long-Term-Care Insurance.” Journal of PoliticalEconomy, 98(1): 153–168.
20
Pool, Lindsay R., Sarah A. Burgard, Belinda L. Needham, Michael R. Elliott, Kenneth M. Langa, andCarlos F. Mendes de Leon. 2018. “Association of a Negative Wealth Shock with All-Cause Mortalityin Middle-Aged and Older Adults in the United States.” Journal of the American Medical Association,319(13): 1341–1350.
Prinz, Daniel, Michael Chernew, David Cutler, and Austin Frakt. 2018. “Health and Economic Activityover the Life Cycle: Literature Review.” National Bureau of Economic Research Working Paper 24865,Cambridge, MA.
RAND. n.d.. RAND HRS Public Use Datasets. Produced by the RAND Center for the Study of Aging, withfunding from the National Institute on Aging and the Social Security Administration. Santa Monica, CA:RAND.
Rosenzweig, Mark R., and Kenneth I. Wolpin. 1993. “Intergenerational Support and the Life-Cycle In-comes of Young Men and Their Parents.” Journal of Labor Economics, 11(1): 84–112.
Rosenzweig, Mark R., and Kenneth I. Wolpin. 1994. “Parental and Public Transfers to Young Womenand Their Children.” American Economic Review, 84(5): 1195–1212.
Salm, Martin. 2009. “Does Job Loss Cause Ill Health?” Health Economics, 18(9): 1075–1089.
Schwandt, Hannes. 2018. “Wealth Shocks and Health Outcomes: Evidence from Stock Market Fluctua-tions.” American Economic Journal: Applied Economics, 10(4): 349–377.
Sharma, Andy. 2015. “Divorce/Separation in Later Life: A Fixed Effects Analysis of Economic Well-Beingby Gender.” Journal of Family and Economic Issues, 36(2): 299–306.
Sloan, Frank A, Harold H Zhang, and Jingshu Wang. 2002. “Upstream Intergenerational Transfers.”Southern Economic Journal, 69(2): 363–380.
Stokes, Jeffrey E., and Sarah E. Patterson. 2020. “Intergenerational Relationships, Family CaregivingPolicy, and COVID-19 in the United States.” Journal of Aging & Social Policy, 32(4-5): 416–424. PMID:32489144.
Wolff, Jennifer L, Brenda C Spillman, Vicki A Freedman, and Judith D Kasper. 2016. “A nationalprofile of family and unpaid caregivers who assist older adults with health care activities.” JAMA internalmedicine, 176(3): 372–379.
Wu, Stephen. 2003. “The Effects of Health Events on the Economic Status of Married Couples.” Journalof Human Resources, 38(1): 219–230.
21
Figure 1: Financial Shocks in the 2010 HRS and Intergenerational Financial Transfers
Panel A: Wealth Loss
-.1-.0
50
.05
7 8 9 10 11 12HRS Wave
No Wave 10 Shock Wave 10 Shock
Parent to Child Transfer
-.1-.0
50
.05
7 8 9 10 11 12HRS Wave
No Wave 10 Shock Wave 10 Shock
Child to Parent Transfer
Panel B: Job Exit
-.1-.0
50
.05
7 8 9 10 11 12HRS Wave
No Wave 10 Shock Wave 10 Shock
Parent to Child Transfer-.1
-.05
0.0
5
7 8 9 10 11 12HRS Wave
No Wave 10 Shock Wave 10 Shock
Child to Parent Transfer
Notes: This figure compares average (unadjusted) transfer rates in each HRS wave among households thatexperienced wealth loss or job exit during wave 10 of the HRS to average transfer rates among householdsthat were present for wave 10 but did not experience a shock. In order to compare trends, each line isnormalized to show changes from wave 8 values. The sample is restricted to households with membersaged 50-85. Observations are weighted using HRS sample weights.
22
Figure 2: Event-Studies: Financial Shocks and Intergenerational Financial Transfers
Panel A: Wealth Loss
-.1-.0
50
.05
-3 -2 -1 0 1 2 3Waves Since Event Recorded
Parent to Child Transfer
-.05
0.0
5.1
-3 -2 -1 0 1 2 3Waves Since Event Recorded
Child to Parent Transfer
Panel B: Job Exit
-.1-.0
50
.05
-3 -2 -1 0 1 2 3Waves Since Event Recorded
Parent to Child Transfer
-.05
0.0
5.1
-3 -2 -1 0 1 2 3Waves Since Event Recorded
Child to Parent Transfer
Notes: This figure shows event-study coefficients and confidence intervals from estimation of Equation 1.Each panel presents a different shock, and each figure presents a different outcome variable. The sample isrestricted to households with members aged 50-85. Standard errors are clustered by household. Observa-tions are weighted using HRS sample weights.
23
Figure 3: Event-Studies: Fatal Health Shocks and Intergenerational Financial Transfers
Panel A: Male Widowed
-.05
0.0
5.1
.15
.2
-3 -2 -1 0 1 2 3Waves Since Event Recorded
Parent to Child Transfer
-.05
0.0
5.1
-3 -2 -1 0 1 2 3Waves Since Event Recorded
Child to Parent Transfer
-.05
0.0
5.1
-3 -2 -1 0 1 2 3Waves Since Event Recorded
Child Helping
Panel B: Female Widowed
-.05
0.0
5.1
.15
.2
-3 -2 -1 0 1 2 3Waves Since Event Recorded
Parent to Child Transfer-.0
50
.05
.1
-3 -2 -1 0 1 2 3Waves Since Event Recorded
Child to Parent Transfer
-.05
0.0
5.1
-3 -2 -1 0 1 2 3Waves Since Event Recorded
Child Helping
Notes: This figure shows event-study coefficients and confidence intervals from estimation of Equation 1. Eachpanel presents a different shock, and each figure presents a different outcome variable. The sample is restrictedto households with members aged 50-85. Standard errors are clustered by household. Observations are weightedusing HRS sample weights.
24
Figure 4: Event-Studies: Health Shocks and Intergenerational Transfers
Panel A: Hospitalization
-.1-.0
50
.05
-3 -2 -1 0 1 2 3Waves Since Event Recorded
Parent to Child Transfer
-.05
0.0
5.1
-3 -2 -1 0 1 2 3Waves Since Event Recorded
Child to Parent Transfer
-.05
0.0
5.1
-3 -2 -1 0 1 2 3Waves Since Event Recorded
Child Helping
Panel B: Disability
-.1-.0
50
.05
-3 -2 -1 0 1 2 3Waves Since Event Recorded
Parent to Child Transfer
-.05
0.0
5.1
-3 -2 -1 0 1 2 3Waves Since Event Recorded
Child to Parent Transfer
-.05
0.0
5.1
-3 -2 -1 0 1 2 3Waves Since Event Recorded
Child Helping
Panel C: Poor Health
-.05
0.0
5.1
-3 -2 -1 0 1 2 3Waves Since Event Recorded
Parent to Child Transfer
-.05
0.0
5.1
-3 -2 -1 0 1 2 3Waves Since Event Recorded
Child to Parent Transfer-.0
50
.05
.1
-3 -2 -1 0 1 2 3Waves Since Event Recorded
Child Helping
Notes: This figure shows event-study coefficients and confidence intervals from estimation of Equation 1. Eachpanel presents a different shock, and each figure presents a different outcome variable. The sample is restrictedto households with members aged 50-85. Standard errors are clustered by household. Observations are weightedusing HRS sample weights.
25
Table 1: Weighted Sample Means by Transfer Status
Full Sample Giving $ Receiving $ Receiving HelpOldest age 65.99 64.01 67.67 71.79Any non-white 0.17 0.14 0.31 0.29Highest education 13.11 14.09 12.20 10.87Real per capita income 48031.29 67060.95 27561.61 23512.25Real per capita assets 297960.50 418722.11 139093.32 135165.11Probability of bequest 43.26 55.92 25.53 18.60Any employed 0.52 0.64 0.41 0.10Home owner 0.78 0.85 0.65 0.53SSI receipt 0.02 0.01 0.04 0.09Hospitalization 0.33 0.31 0.39 0.55Any disability 0.24 0.19 0.36 0.90Any bad health 0.35 0.28 0.47 0.76Out-of-pocket medical expenses 5534.91 6110.61 5501.18 8513.55Life expectancy 58.80 62.95 53.97 38.26Single male 0.10 0.11 0.06 0.11Single female 0.31 0.23 0.58 0.69Number of children 3.36 3.15 3.81 4.12Any child within 10 miles 0.47 0.46 0.48 0.57Observations 113781 36079 6594 7082Notes: The data are from waves 2 through 12 of the Health and Retirement Study (HRS). Columns2–4 show average outcomes for household-wave observations that report giving or receiving trans-fers. Real income, wealth, and medical expenses are reported in 2017 dollars. The sample is re-stricted to households with members aged 50–85. Observations are weighted using HRS sampleweights.
26
Table 2: Intergenerational Transfers in HRS Households
Panel A: Full Sample and by Age, Wealth, and EmploymentOldest Age Wealth Percentile
Full Sample 75 to 85 50 to 60 Bottom 20 Top 20 EmployedParent-to-child transfer 0.35 0.27 0.45 0.19 0.48 0.43Parent-to-child amt, nonzero 13560.82 16388.27 13147.54 6118.65 24165.18 13478.14Child-to-parent transfer 0.05 0.06 0.04 0.09 0.02 0.04Child-to-parent amt, nonzero 5013.49 5757.52 4466.33 4046.39 9258.50 4710.66Probability of bequest 43.26 40.16 44.63 8.67 79.32 48.79ADL/IADL help 0.05 0.11 0.03 0.12 0.02 0.01Observations 113781 31689 30790 22789 22756 51550
Panel B: By Household Structure and Health StatusHousehold Structure Health Status
Female Male Couple Poor Excellent DisabledParent-to-child transfer 0.27 0.36 0.40 0.21 0.45 0.27Parent-to-child amt, nonzero 10873.83 15229.47 14241.02 10988.73 18885.14 10939.57Child-to-parent transfer 0.09 0.03 0.03 0.08 0.03 0.07Child-to-parent amt, nonzero 4726.00 4637.44 5540.76 4719.74 6425.52 5005.15Probability of bequest 29.78 39.92 50.76 18.37 61.06 29.41ADL/IADL help 0.11 0.06 0.02 0.24 0.01 0.19Observations 40010 10850 62921 9794 12944 30538Notes: Real transfer amounts are reported in 2017 dollars. The sample is restricted to households with membersaged 50–85. Observations are weighted using HRS sample weights.
27
Table 3: Cross-Sectional and Fixed Effects Regressions
Event: Giving $ Receiving $ Receiving helpCS FE CS FE CS FE
Wealth Q2 0.052 0.039 -0.005 -0.004 -0.021 -0.019(0.006) (0.007) (0.004) (0.004) (0.004) (0.004)
Wealth Q3 0.080 0.039 -0.014 -0.007 -0.016 -0.027(0.007) (0.008) (0.004) (0.005) (0.004) (0.004)
Wealth Q4 0.116 0.058 -0.024 -0.007 -0.017 -0.031(0.008) (0.009) (0.004) (0.005) (0.004) (0.004)
Wealth Q5 0.149 0.070 -0.031 -0.004 -0.019 -0.037(0.009) (0.011) (0.004) (0.005) (0.004) (0.005)
Income Q2 0.047 0.007 0.000 0.001 -0.010 0.001(0.005) (0.006) (0.004) (0.003) (0.004) (0.003)
Income Q3 0.108 0.040 -0.017 -0.010 -0.019 -0.005(0.006) (0.006) (0.003) (0.004) (0.004) (0.003)
Income Q4 0.152 0.047 -0.025 -0.013 -0.010 0.003(0.007) (0.007) (0.004) (0.004) (0.003) (0.003)
Income Q5 0.241 0.093 -0.030 -0.016 0.001 0.008(0.008) (0.008) (0.004) (0.004) (0.003) (0.003)
Any employed 0.004 0.002 0.012 0.002 -0.017 -0.003(0.006) (0.006) (0.002) (0.003) (0.002) (0.002)
Single male 0.024 0.071 -0.003 0.005 0.022 0.056(0.010) (0.014) (0.003) (0.005) (0.003) (0.005)
Single female -0.023 0.003 0.054 0.038 0.071 0.071(0.006) (0.010) (0.003) (0.005) (0.003) (0.005)
Hospitalization 0.011 0.003 0.006 0.007 0.014 0.010(0.004) (0.004) (0.002) (0.002) (0.002) (0.002)
Any disability -0.000 -0.001 0.015 0.014 0.149 0.115(0.005) (0.005) (0.002) (0.002) (0.003) (0.003)
Any bad health 0.006 0.002 0.004 0.002 0.017 0.010(0.005) (0.005) (0.002) (0.002) (0.002) (0.002)
Mean 0.317 0.317 0.058 0.058 0.062 0.062Notes: All regressions include a quadratic in age, an indicator for nonwhite re-spondent or spouse, indicators for highest educational attainment (high schoolgraduate, college graduate), number of children, and a full set of HRS cohort-by-wave dummies.
28
Table 4: Event-Study: Parental Household Events and Intergenerational Transfers
Wealth Job Male Female Any Disability PoorEvent: loss exit widowed widowed hosp onset health
Panel A: Any Parent-to-Child TransferEvent Wave -0.012 -0.013 0.126 0.050 -0.019 -0.011 -0.009
(0.007) (0.007) (0.020) (0.014) (0.008) (0.009) (0.015)One Wave After -0.034 -0.034 0.081 0.023 -0.021 -0.010 -0.024
(0.009) (0.008) (0.026) (0.016) (0.009) (0.011) (0.019)
Mean 0.411 0.439 0.337 0.363 0.393 0.369 0.353Panel B: Any Child-to-Parent Transfer
Event Wave 0.008 0.008 0.022 0.021 0.003 0.013 0.017(0.003) (0.003) (0.012) (0.007) (0.003) (0.004) (0.007)
One Wave After 0.011 0.002 -0.002 0.007 -0.002 0.016 0.016(0.004) (0.004) (0.012) (0.008) (0.004) (0.006) (0.009)
Mean 0.040 0.040 0.038 0.041 0.044 0.050 0.046Panel C: Any Child Helps with ADLs
Event Wave 0.020 0.008 0.082 0.069 0.024 0.106 0.081(0.003) (0.003) (0.012) (0.008) (0.003) (0.005) (0.009)
One Wave After 0.023 0.011 0.055 0.061 0.021 0.082 0.057(0.003) (0.003) (0.012) (0.008) (0.004) (0.006) (0.011)
Mean 0.023 0.009 0.012 0.021 0.024 0.016 0.035Observations 68217 64269 6676 22147 61535 42476 13524Households 9253 7949 894 2722 7372 5149 1672Notes: This table presents event-study coefficients and confidence intervals from estimationof Equation 2. Each panel presents a different shock, and each column presents a differ-ent outcome variable. The sample is restricted to households with members aged 50-85.Standard errors are clustered by household. Observations are weighted using HRS sampleweights.
29
Table 5: Event-Study: Specific Health Diagnoses and Intergenerational Transfers
Event: Cardiac Stroke Memory Cancer
Panel A: Any Parent-to-Child TransferEvent Wave -0.003 -0.017 0.011 -0.023
(0.012) (0.018) (0.019) (0.016)One Wave After -0.019 -0.029 -0.020 -0.051
(0.014) (0.021) (0.023) (0.019)
Mean 0.392 0.363 0.296 0.409Panel B: Any Child-to-Parent Transfer
Event Wave 0.001 -0.002 0.011 0.001(0.006) (0.009) (0.012) (0.007)
One Wave After 0.003 0.016 0.004 -0.006(0.007) (0.011) (0.014) (0.007)
Mean 0.045 0.045 0.062 0.038Panel C: Any Child Helps with ADLs
Event Wave 0.024 0.061 0.080 0.018(0.005) (0.011) (0.014) (0.006)
One Wave After 0.025 0.055 0.086 0.019(0.007) (0.012) (0.017) (0.006)
Mean 0.027 0.037 0.067 0.018Observations 24498 9753 7428 15321Households 2701 1087 859 1656Notes: This table presents event-study coefficients andconfidence intervals from estimation of Equation 2. Eachpanel presents a different shock, and each column presentsa different outcome variable. The sample is restricted tohouseholds with members aged 50-85. Standard errors areclustered by household. Observations are weighted usingHRS sample weights.
30
Table 6: Mechanisms
Wealth Job Male Female Any Disability PoorEvent: loss exit widowed widowed hosp onset health
Panel A: Wealth per Capita ($1000s)One Wave After -181.95 13.63 164.14 168.02 -19.30 -7.53 26.58
(11.36) (10.42) (25.16) (21.13) (10.64) (13.21) (25.35)
Mean 361.69 292.13 261.27 259.45 303.58 273.06 248.49Panel B: Probability of Leaving a Bequest of at Least $100,000
One Wave After -4.95 0.06 7.16 1.60 -0.96 -0.55 0.68(0.63) (0.58) (2.06) (1.13) (0.66) (0.80) (1.54)
Mean 45.71 45.05 45.27 40.71 44.31 40.07 36.18Panel C: Income per Capita ($1000s)
One Wave After -2.32 -13.39 10.15 11.01 -5.39 1.06 2.06(1.73) (1.42) (6.41) (3.10) (1.53) (2.22) (1.99)
Mean 53.68 61.43 41.21 37.96 51.41 45.15 38.70Panel D: Out-of-Pocket Medical Expenses ($1000s)
One Wave After -0.58 0.51 -7.49 -5.24 1.58 2.12 1.20(0.26) (0.31) (0.83) (0.63) (0.26) (0.43) (0.65)
Mean 5.56 5.27 8.25 7.69 4.75 5.49 6.36Panel E: Life Expectancy (probability)
One Wave After -1.44 0.13 1.48 -1.75 -1.66 -2.36 -3.73(0.50) (0.43) (1.73) (0.97) (0.52) (0.66) (1.22)
Mean 62.91 65.90 53.74 60.36 63.73 60.33 58.24
Notes: This table presents event-study coefficients and confidence intervals from estimationof Equation 2. Each panel presents a different outcome variable, and each column presents adifferent shock. The sample is restricted to households with members aged 50-85. Standarderrors are clustered by household. Observations are weighted using HRS sample weights.
31
Table 7: Parental Household Events and Intergenerational Transfers, by Wealth
Wealth Job Male Female Any Disability PoorEvent: loss exit widowed widowed hosp onset health
Panel A: Any Parent-to-Child TransferLow Wealth
Event Wave -0.019 -0.013 0.100 0.063 -0.022 -0.011 -0.031(0.011) (0.010) (0.027) (0.020) (0.011) (0.011) (0.020)
One Wave After -0.061 -0.044 0.049 0.045 -0.022 -0.013 -0.052(0.013) (0.012) (0.039) (0.023) (0.014) (0.015) (0.024)
Mean 0.326 0.347 0.256 0.268 0.292 0.275 0.277High Wealth
Event Wave -0.010 -0.013 0.148 0.040 -0.018 -0.012 0.017(0.010) (0.010) (0.030) (0.019) (0.010) (0.013) (0.022)
One Wave After -0.016 -0.027 0.106 0.008 -0.020 -0.008 0.005(0.013) (0.011) (0.035) (0.023) (0.013) (0.016) (0.029)
Mean 0.473 0.510 0.405 0.428 0.467 0.450 0.438Panel B: Any Child-to-Parent Transfer
Low WealthEvent Wave 0.021 0.017 0.036 0.044 0.010 0.018 0.025
(0.006) (0.006) (0.021) (0.014) (0.007) (0.007) (0.012)One Wave After 0.022 0.003 -0.009 0.014 -0.001 0.026 0.023
(0.007) (0.007) (0.019) (0.016) (0.008) (0.010) (0.016)
Mean 0.062 0.062 0.061 0.063 0.069 0.076 0.066High Wealth
Event Wave -0.003 0.001 0.008 0.006 -0.002 0.008 0.007(0.003) (0.003) (0.012) (0.008) (0.003) (0.005) (0.008)
One Wave After 0.001 0.000 -0.000 0.004 -0.002 0.006 0.006(0.004) (0.004) (0.014) (0.008) (0.004) (0.006) (0.010)
Mean 0.024 0.022 0.019 0.026 0.025 0.028 0.023Panel C: Any Child Helps with ADLs
Low WealthEvent Wave 0.025 0.018 0.106 0.106 0.033 0.128 0.076
(0.005) (0.005) (0.020) (0.014) (0.006) (0.008) (0.013)One Wave After 0.031 0.020 0.076 0.093 0.032 0.102 0.064
(0.006) (0.006) (0.021) (0.015) (0.007) (0.009) (0.015)
Mean 0.035 0.015 0.021 0.035 0.044 0.024 0.047High Wealth
Event Wave 0.013 0.001 0.064 0.045 0.018 0.084 0.086(0.003) (0.003) (0.013) (0.008) (0.003) (0.007) (0.012)
One Wave After 0.014 0.003 0.042 0.042 0.012 0.063 0.048(0.004) (0.004) (0.014) (0.009) (0.004) (0.007) (0.014)
Mean 0.014 0.004 0.004 0.012 0.010 0.009 0.021Notes: See Table 4 notes.
Table A1: Specific Non-Fatal Health Diagnoses: Mechanisms
Event: Cardiac Stroke Memory Cancer
Panel A: Wealth per Capita ($1000s)One Wave After -8.28 19.58 56.92 33.93
(15.52) (32.52) (39.56) (27.42)
Mean 295.82 265.89 232.92 319.81Panel B: Probability of Leaving a Bequest of at Least $100,000
One Wave After -1.09 1.33 2.23 -0.57(1.09) (1.70) (1.83) (1.34)
Mean 44.19 38.48 33.63 47.54Panel C: Income per Capita ($1000s)
One Wave After 1.32 -0.38 -0.52 -1.55(2.33) (2.43) (3.01) (1.91)
Mean 48.87 45.03 37.45 52.17Panel D: Out-of-Pocket Medical Expenses ($1000s)
One Wave After 2.08 2.94 1.92 1.65(0.40) (0.82) (1.38) (0.50)
Mean 5.13 5.90 7.09 5.54Panel E: Life Expectancy (probability)
One Wave After -2.54 -2.36 -0.74 -2.02(0.82) (1.42) (1.78) (0.98)
Mean 63.76 60.43 57.83 64.31
Notes: See Table 6 notes.
33
Table A2: Mechanisms - Low Wealth Households
Wealth Job Male Female Any Disability PoorEvent: loss exit widowed widowed hosp onset health
Panel A: Wealth per Capita ($1000s)One Wave After -49.63 -0.87 25.32 23.29 -5.52 -3.52 -1.05
(2.00) (1.87) (5.95) (3.77) (1.89) (2.28) (3.19)
Mean 64.00 46.79 37.05 45.75 46.72 45.12 46.71Panel B: Probability of Leaving a Bequest of at Least $100,000
One Wave After -5.46 -1.31 1.38 -3.58 -1.29 0.06 -0.70(0.95) (0.86) (2.94) (1.45) (0.97) (1.08) (1.98)
Mean 21.08 20.55 17.45 14.47 17.65 15.33 15.15Panel C: Income per Capita ($1000s)
One Wave After -3.71 -11.23 8.30 3.62 -2.80 -3.59 2.85(1.94) (2.33) (1.62) (1.12) (0.78) (2.90) (1.94)
Mean 33.89 38.15 23.35 23.52 30.19 28.58 24.97Panel D: Out-of-Pocket Medical Expenses ($1000s)
One Wave After -0.91 0.10 -6.96 -5.38 1.49 1.72 0.91(0.43) (0.38) (1.35) (1.06) (0.47) (0.55) (0.88)
Mean 5.28 5.26 8.53 7.22 4.27 4.72 5.88Panel E: Life Expectancy (probability)
One Wave After -0.53 -1.19 2.13 0.46 -1.38 -3.60 -3.89(0.83) (0.72) (2.81) (1.73) (0.91) (1.01) (1.79)
Mean 58.31 60.87 51.55 55.41 58.95 57.59 56.05
Notes: See Table 6 notes.
34
Table A3: Mechanisms - High Wealth Households
Wealth Job Male Female Any Disability PoorEvent: loss exit widowed widowed hosp onset health
Panel A: Wealth per Capita ($1000s)One Wave After -275.69 23.61 266.92 256.52 -25.41 -5.04 62.07
(19.99) (17.83) (42.92) (34.31) (18.10) (24.50) (52.95)
Mean 576.97 479.70 452.14 404.01 491.06 470.12 473.71Panel B: Probability of Leaving a Bequest of at Least $100,000
One Wave After -3.90 1.02 11.03 4.62 -0.60 -0.83 2.23(0.87) (0.79) (2.78) (1.60) (0.90) (1.15) (2.40)
Mean 63.28 63.68 66.77 58.25 63.41 61.15 58.91Panel C: Income per Capita ($1000s)
One Wave After -2.68 -15.09 11.68 15.66 -7.45 4.42 1.13(2.57) (1.84) (11.59) (5.06) (2.59) (3.26) (3.61)
Mean 67.99 79.23 56.41 47.73 66.89 59.48 54.03Panel D: Out-of-Pocket Medical Expenses ($1000s)
One Wave After -0.21 0.82 -7.99 -5.19 1.67 2.57 1.66(0.32) (0.45) (1.03) (0.78) (0.31) (0.63) (0.95)
Mean 5.76 5.27 8.02 8.01 5.09 6.15 6.90Panel E: Life Expectancy (probability)
One Wave After -2.00 1.03 1.33 -2.92 -1.82 -1.28 -3.96(0.63) (0.54) (2.19) (1.15) (0.61) (0.86) (1.66)
Mean 66.12 69.62 55.44 63.57 67.03 62.57 60.54
Notes: See Table 6 notes.
35